The identification of potential business failure is important to many corporate managers, investors, bankers and auditors. If the warning signs exhibited ahead of failure can be identified in advance, then appropriate action can be undertaken hopefully to reverse the process or at least to minimise the possible damage. One potential way to improve such judgements is to improve the information presentation format.

Bankruptcy prediction is typically a multivariate decision requiring the consideration of multiple financial variables over multiple time periods. Since multivariate graphics provide spatial integration, summarization and comparison of financial variables, they may result in better decisions.

The traditional and more familiar tabular presentation of data is typically available to decision makers, though they can usually refer to specific data values before making a final decision. The potential for improved decision making exists if additional information can be gleaned from the presentation of data in alternative formats.

Given that tabular presentations are familiar, useful and normally available, this study tests for the incremental benefit of multivariate graphics over tabular formats (i.e., by comparing tabular–graphical combination formats with a tabular-alone format), an area where research efforts have been sparse. Even then results have been inconsistent (as detailed below).

In addition to the more traditional multivariate graphic presentations of bar charts, this study includes the more innovative form of schematic faces developed by Chernoff (1973). By assigning variables and their values to facial features, changes in expression, can provide a quick indication of the relationship among variables. A single global judgement is promptly facilitated despite the novelty of the approach.

Because task characteristics (essentially task type and task complexity) have been shown to be important in decision performance across different presentation formats (see reviews and meta-analyses: DeSanctis, 1984; Hwang & Wu, 1990; Jarvenpaa & Dickson, 1988; Montazemi & Wang, 1988–1989; and other works: Amer, 1991; Blocher, Moffie, & Zmud, 1986; Davis, 1989 and Hard and Vanecek, 1991; Umanath & Vessey, 1994; Wright, 1995) report format and task characteristics must be considered in an interactive manner in examining decision performance across different formats.

The issue of task characteristics is, however, quite complex, given the variety of definition, interpretation and measurement, and the absence of a ready taxonomy of classification. Task characteristics have many dimensions, among them, those more commonly reported in the literature: task type and task complexity. Task complexity can be further distinguished into information complexity and job complexity. Such a distinction, however, is rarely made in the literature, further complicating this issue because of alternative ways of measuring information complexity and job complexity.

The current study focuses on the multivariate decision task and on information complexity, because information complexity is more fundamental than job complexity and can be objectively assessed, independent of any particular task-doer, and unaffected by the presentation format. The current study also initiates a novel method to measure information complexity, based on the internal consistency of the information, developed by So and Smith (2002) but originally suggested by Steinmann (1976). This is of particular relevance to the multivariate decision tasks of this study.

This study seeks to determine the effective presentation format (tabular–graphical combination, tabular–facial combination, or tabular-alone format) for the performance of multivariate decision tasks of varying information complexity, through an experimental study which examines the interactive influence of presentation format and information complexity on multivariate decision performance.

A number of accounting studies, addressing the presentation of multivariate financial information, compare tabular presentation with multivariate graphics, including bar charts, line graphs and schematic faces. The general observation is that multidimensional graphics take less time to finish, and are more accurate than tabular presentations, though much of the evidence for the latter is inconsistent.

Studies addressing the incremental usefulness of multivariate graphics over tables are rare, and are restricted to comparisons of tabular presentations with tabular presentations supplemented by graphs. Benbasat and Dexter (1986) ask graduate students to allocate monetary resources across various income generating alternatives, given tabular, graphical, and combined tabular and graphical presentations of profit relationships. Their results provide evidence that the tabular format leads to better decision making while the graphical format leads to faster decision making. A combined format, which integrates the advantages associated with both tabular and graphical presentation, is found to be the superior format (in both profit and time performance) and is rated very highly by decision makers.

Results from DeSanctis and Jarvenpaa (1989) also provide support, though not as strong as in Benbasat and Dexter (1986), for the contention that combined graphical/numeric formats provide incremental value over numeric formats and graphical formats in forecasting financial statement information. DeSanctis and Jarvenpaa (1989) find such incremental value to occur only after practice has been provided to subjects. The combined group has the highest decision accuracy of the three groups and gradually improves with successive trials. While confidence is also the highest in the combined group, the difference in confidence ratings is not statistically significant.

Nibbelin et al. (1992) extend the prior studies by including additional presentation formats while also considering the individual characteristic of field dependence. They examine tabular presentations supplemented by schematic faces and tabular presentations supplemented by bar charts, in a bond-rating-change decision. Their results indicate that the group receiving only tabular data achieves the highest accuracy, those receiving supplemental bar graphs the second highest, and those receiving supplemental faces the lowest. Thus the advantages of faces claimed in earlier studies on a ‘faces-alone’ format are not supported. The group receiving supplemental faces is, however, significantly faster than either of the other groups. Field dependence is found to be a significant determinant of accuracy, but not of decision time. Subjects who are more field independent (i.e., those more able to provide organization to a disorganized field of information and thus sample more fully from the clues that are available) perform more accurately at the bond-rating task. Field dependence is, however, not useful in explaining the differences in decision accuracy among the different formats.

Wright (1995) considers three levels of task complexity (i.e., three levels of information acquisition and mental integration) and finds that the incremental availability of graphs results in better judgement by auditors, but only when the task is complex. When the number of relevant information relationships is low and the required information integration is simple (e.g., in the judgement of short-term liquidity) very little or no incremental benefit from graphs is apparent. In contrast, when auditors are required to provide comprehensive and demanding loan collectibility judgements, the availability of graphical results helps to eliminate bias towards an overestimation of loan collectibility, and to reduce judgement error.